Alignment Midtraining for Animals

Research project and 2026 preprint by CaML that uses synthetic midtraining documents to improve language models’ compassionate reasoning about animals and introduces the ANIMA benchmark for evaluating animal-welfare reasoning.

Endorsements support Compassion in Machine Learning.

No endorsements yet.

Discussion

Start Date

-

End Date

-

Expected Duration

-

Funding Raised to Date

-